Model-based cardiovascular parameter estimation in the Intensive Care Unit

In this paper we present a simulation study that aims at estimating parameters of a hemodynamic model using observable data typically available in an Intensive Care Unit (ICU). Tracking model parameters in time reveals disease progression, and hence can be very useful for patient monitoring purposes. However, the observable data is generally not rich enough to allow for reliable estimation of all parameters of the underlying model. This leads to an 'ill-conditioned' estimation problem. To overcome this ill-conditioning, we employ subset selection to identify the 'well-conditioned' parameters that can be estimated robustly. We attempt to estimate only these parameters while the rest are fixed at prior values. Our results indicate that focusing on the reduced-order estimation problem improves the reliability of the estimates by more than 50%; the scheme is capable of recovering the underlying well-conditioned parameters with reasonable accuracy in both steady-state and transient conditions.